<FLD>
has anyone looked into removing the verification from tp-links u-boot bootloader so it can be binary patched?
<Namidairo>
if it's spitting out an error on checksum failure you should be able to follow it back...
rsalvaterra has quit [Quit: rsalvaterra]
<FLD>
follow it back?
rsalvaterra has joined #openwrt-devel
<FLD>
now that i'm looking at the GPL source for the uboot, i noticed there's MAKEOPTS_COMPRESSED_UBOOT=1 so i guess that makes it a moot point anyways
<FLD>
only way to modify the uboot would be then to somehow get it recompiled from the GPL source
danitool has quit [Ping timeout: 480 seconds]
<hurricos>
hey, does anyone have experience tracing LEDs?
<hurricos>
I'm trying to figure out how to configure (with which GPIO regiser offsets) the device tree
<hurricos>
so that the device tree actually properly maps LEDs to aliases
<hurricos>
and I'm wondering if there's a way to "raw" trigger GPIOs
<neggles>
mrkiko: as far as I can tell, 1.2GHz max clock - the flow-offload engines might be entirely different though. it's still 2c4t mips32r2, built in DBDC 802.11ac
<mrkiko>
neggles: mtk based wifi ?? Or siflower-based or different?
<neggles>
mrkiko: it's a binary blob driver with its own module, but from a bit of a poke around in the datasheet the registers/config seem to be, shall we say, "inspired by" mtk
<mrkiko>
neggles: thanks
<neggles>
this kernel has debugfs enabled and a whole bunch of stuff exposed
<neggles>
ooooh, that's a major improvement over the MT7621 - the USB-OTG controller is a DWC2
<neggles>
they've done a remarkably good job with this openwrt tree as well, even if it is 18.06 / stuck on 4.14 kernel - they've made themselves a target, put the binary kernel modules into packages instead of the kernel source tree... a vast improvement on the usual hackjob
* enyc
meows
victhor has joined #openwrt-devel
<Namidairo>
at least they have cfg80211 support
<Namidairo>
lol their kernel modules are marked GPL in modinfo
<nitroshift>
stintel, rsalvaterra i just ordered a mt7615-basec m.2 card off asiarf.com
<Namidairo>
there are a bunch of suspicious looking listings for mt7921k cards now
<Namidairo>
with amd pci ids
<nitroshift>
rsalvaterra, 1733 is enough for me transfer-wise
<rsalvaterra>
nitroshift: Sure, but I'd rather maximise the system spectral efficiency. Peak throughput in a couple of devices isn't that interesting to me. :)
<f00b4r0>
i still believe in the holy appendage. Wire forever :)
<rsalvaterra>
f00b4r0: Amen!
<nitroshift>
f00b4r0, me too but i can't wire the tv, wife will kill me (literally!)
<f00b4r0>
lol
<Namidairo>
with the wire
<nitroshift>
Namidairo, she'll use the wire to hang me by my feet
<rsalvaterra>
What's with women and wiring, anyway? :P
* rsalvaterra
has been there too.
danitool has quit [Quit: Cubum autem in duos cubos, aut quadratoquadratum in duos quadratoquadratos]
<f00b4r0>
urg! Dist-upgraded and now spamassassin kills legit mail.
<rsalvaterra>
Good God, the packaging of qdisc/act/match modules is a real mess.
* rsalvaterra
wants to have the minimal set of functionality required by qosify.
<rsalvaterra>
I believe those should be the contents of kmod-sched-core. Nothing more.
<rsalvaterra>
(Well, the BPF cls/act can be a separate package, like it is today.)
ashkan has joined #openwrt-devel
<ashkan>
does anyone have a working branch of x86 with kernel 5.15 ?
<ashkan>
plz :D
KGB-0 has quit [Ping timeout: 480 seconds]
Tapper has quit [Ping timeout: 480 seconds]
<nitroshift>
ashkan, not yet
<nitroshift>
it fails to build hv_util
<nitroshift>
unless someone is willing to look into it
minimal has joined #openwrt-devel
<stintel>
rsalvaterra: told you so :P
<rsalvaterra>
stintel: I'm a stubborn bastard. :P
danitool has joined #openwrt-devel
KGB-1 has quit [Remote host closed the connection]
KB1SPH has joined #openwrt-devel
nitroshift has quit [Quit: Gone that way --->]
valku has joined #openwrt-devel
Tapper has joined #openwrt-devel
<rsalvaterra>
stintel: Nevertheless, while investigating, I found two related fixes which seem beneficial by themselves (patches sent).
<stintel>
rsalvaterra: nice
<stintel>
hauke: surely you meant version 20211216 instead of 20121216 ? :)
EqUaTe_ has joined #openwrt-devel
<rsalvaterra>
Endian slip? :)
<stintel>
or nostradamus correction? :P
EqUaTe has quit [Read error: Connection reset by peer]
EqUaTe_ is now known as EqUaTe
KGB-0 has joined #openwrt-devel
bluew has quit [Remote host closed the connection]
bluew has joined #openwrt-devel
<Grommish>
*sigh* -- Build completed unsuccessfully in 7:51:26
<Grommish>
Meh
<stintel>
:D
<stintel>
rust ? :)
<Grommish>
stintel: Yeah. I'm working on native binaries for the devices
<Grommish>
just takes so bloody long to test
<stintel>
yeah, one more reason to extract the llvm part out of the rust package I guess
<Grommish>
stintel: Oh, this isn't because of LLVM, although that can take a while first time thru
<Grommish>
stintel: It's just everything else
<Grommish>
stintel: LLVM actually has caused very few issues so far.. PPC is one, but *shrug* other than that, it's been ok Ive had issues with just about everything else though, inluding the rust libc
<Grommish>
stintel: I thought about trying to re-purpose the LLVM bpf stuff, but i think that'll just add far more complexity
Tapper has quit [Read error: Connection reset by peer]
Tapper has joined #openwrt-devel
kenny has joined #openwrt-devel
Tapper has quit [Ping timeout: 480 seconds]
goliath has quit [Quit: SIGSEGV]
<stintel>
rsalvaterra: if you look a bit deeper at the 2 commits that I just pushed you'll understand my frustration too :P
danitool has quit [Quit: Cubum autem in duos cubos, aut quadratoquadratum in duos quadratoquadratos]
<stintel>
alright
<stintel>
qoriq v2 coming up
cmonroe has joined #openwrt-devel
<hurricos>
stintel: you rock
<hurricos>
neggles: you're right, gpio expanders aren't controlled by GPIO. I just gotta crack the little thing open and look :^)
<hurricos>
> siflower MT7621 < :eyes:
<\x>
hows that clone? is it better? core clockspeed seems better atleast I guess
<stintel>
hurricos: I wish
<stintel>
hurricos: if anything, I trance :P
<\x>
not a lot of mt7621 will handle 1200MHz overclock, most does like 1100~1160
<\x>
im speaking from experience here
<stintel>
hope I didn't screw anything up
<stintel>
fucking hell
<stintel>
fucking netifd
<stintel>
setting mtu 1386 for no apparent reason, fucking everything up (dynamic routing does not like mtu mismatch)
<stintel>
undocumented piece of shit
<dhewg>
nbd: on ipq4019 I get "*** NOT YET: opcode 3f ***", it look like an unsupported bpf opcode, so likely due to qosify. Is that expected or a bug?
Tapper has joined #openwrt-devel
aparcar has quit [Read error: Connection reset by peer]
aparcar has joined #openwrt-devel
<nbd>
dhewg: which llvm toolchain did you use?
srslypascal is now known as Guest8959
srslypascal has joined #openwrt-devel
dedeckeh has joined #openwrt-devel
Guest8959 has quit [Ping timeout: 480 seconds]
<dhewg>
clang-13 from the debian build host
<nbd>
does the same happen if you install prebuilt qosify from the snapshot?
<nbd>
the opcode looks weird to me
<dhewg>
dunno, I can try
<dhewg>
I'm not exactly sure when this happens, maybe only on the first wan connection, I'll keep an eye open
<nbd>
if it's qosify, it should happen while the qosify daemon starts
<nbd>
if it happens some other time, e.g. on a first wan connection, it must be something else
<nbd>
since this message is coming from the BPF JIT
<nbd>
and qosify loads its bpf program only once at startup
Tapper has quit [Quit: Tapper]
<dhewg>
hm, i can't force that log output with /etc/init.d/qosify restart
<dhewg>
and not with un/reloading [cls|act]_bpf.ko either
<dhewg>
it seems it's getting printed once upon boot, maybe a red herring, but somewhat around setting up br-lan
<nbd>
could be seccomp related
<nbd>
maybe a bug in procd-ujail
<nbd>
since that one creates some legacy bpf filters
pmelange has joined #openwrt-devel
<dhewg>
hm ok, it's always 3f afaict
GNUmoon has quit [Ping timeout: 480 seconds]
<dhewg>
I had the same ujail/seccomp/qosify setup on mips before (used that .config as start), but I don't think I've seen that there
<dhewg>
maybe some endian thing
<nbd>
i think mips simply doesn't print the warning
<nbd>
so the same bug might be present there too
<dhewg>
ah heh
<nbd>
the warning is coming from the ARM BPF JIT
<nbd>
but it doesn't look like JIT not supporting a particular opcode
<nbd>
the opcode itself seems invalid
ashkan has quit [Quit: Konversation terminated!]
slh64 has quit [Read error: No route to host]
slh has quit [Read error: No route to host]
slh has joined #openwrt-devel
matt__ has joined #openwrt-devel
<dhewg>
it's using pr_info_once though, which explains why I'm seeing it just one time
<dhewg>
might still be qosify?
dedeckeh has quit [Remote host closed the connection]
<nbd>
you could try /etc/init.d/qosify disable and reboot
slh64 has joined #openwrt-devel
<dhewg>
ok, I'll try that tomorrow when the device isn't in use
matt__ has quit [Quit: Leaving]
scappylappy has joined #openwrt-devel
pmelange has left #openwrt-devel [#openwrt-devel]
goliath has joined #openwrt-devel
GNUmoon has joined #openwrt-devel
rua has quit [Ping timeout: 480 seconds]
rua has joined #openwrt-devel
<dhewg>
so everybody disconnected already, and I just tried. The log line doesn't appear then, but if I start qosify later manually, it pops up instantly
<nbd>
ok
<nbd>
then please try reinstalling qosify from the snapshot build
<nbd>
and see if the same appears then
<nbd>
after a reboot
<dhewg>
yeah, happens with those binaries too
<nbd>
ok
<nbd>
ok, checked the disassembly
<nbd>
apparently i was wrong and the opcode is valid
pkgadd_ has joined #openwrt-devel
<nbd>
it's a simple division
pkgadd_ has quit []
rua has quit [Ping timeout: 480 seconds]
<nbd>
BPF_ALU | BPF_DIV | BPF_X == 0x3f
<nbd>
hm, i don't see how it can go from there to the notyet case
<nbd>
you're using kernel 5.10, right?
<dhewg>
yeah
<nbd>
ah, it's BPF_ALU64
<nbd>
so 64-bit division
<nbd>
which the code does not support
<nbd>
so i guess i need to see if i can force the code to make it 32 bit
<dhewg>
at least for 32bit targets
<dhewg>
ppc 32bit doesn't seem to support that opcode too, but mips 32bit does
goliath has quit [Quit: SIGSEGV]
rua has joined #openwrt-devel
* enyc
meeps
* enyc
used to BPF = Berkeley Packet Filter and forgot even when that mattered... =)